A&A, for Press ReleaseDOI: 10.1051/0004-6361:20065940c© ESO 2006

Astronomy&

Astrophysics

3.9 day orbital modulation in the TeV γ-ray flux and spectrumfrom the X-ray binary LS 5039

F. Aharonian1, A. G. Akhperjanian2, A. R. Bazer-Bachi3, M. Beilicke4, W. Benbow1, D. Berge1, K. Bernlöhr1,5,C. Boisson6, O. Bolz1, V. Borrel3, I. Braun1, A. M. Brown7, R. Bühler1, I. Büsching8, S. Carrigan1, P. M. Chadwick7,L.-M. Chounet9, R. Cornils4, L. Costamante1,22, B. Degrange9, H. J. Dickinson7, A. Djannati-Ataï10, L. O’C. Drury11,G. Dubus9, K. Egberts1, D. Emmanoulopoulos12 , P. Espigat10, F. Feinstein13, E. Ferrero12, A. Fiasson13, G. Fontaine9,

Seb. Funk5, S. Funk1, M. Füßling5, Y. A. Gallant13, B. Giebels9, J. F. Glicenstein14, P. Goret14, C. Hadjichristidis7 ,D. Hauser1, M. Hauser12, G. Heinzelmann4, G. Henri15, G. Hermann1, J. A. Hinton1,12, A. Hoffmann16, W. Hofmann1,

M. Holleran8, D. Horns16, A. Jacholkowska13, O. C. de Jager8, E. Kendziorra16, B. Khélifi9,1, Nu. Komin13,A. Konopelko5, K. Kosack1, I. J. Latham7, R. Le Gallou7, A. Lemière10, M. Lemoine-Goumard9, T. Lohse5,

J. M. Martin6, O. Martineau-Huynh17, A. Marcowith3, C. Masterson1,22, G. Maurin10, T. J. L. McComb7, E. Moulin13,M. de Naurois17 ,�, D. Nedbal18, S. J. Nolan7, A. Noutsos7, K. J. Orford7, J. L. Osborne7, M. Ouchrif17,22, M. Panter1,

G. Pelletier15, S. Pita10, G. Pühlhofer12, M. Punch10, B. C. Raubenheimer8, M. Raue4, S. M. Rayner7, A. Reimer19,O. Reimer19, J. Ripken4, L. Rob18, L. Rolland14, G. Rowell1,��, V. Sahakian2, A. Santangelo16, L. Saugé15,

S. Schlenker5, R. Schlickeiser19, R. Schröder19, U. Schwanke5, S. Schwarzburg16, A. Shalchi19, H. Sol6, D. Spangler7,F. Spanier19, R. Steenkamp20, C. Stegmann21, G. Superina9, J.-P. Tavernet17, R. Terrier10, M. Tluczykont9,22,

C. van Eldik1, G. Vasileiadis13, C. Venter8, P. Vincent17, H. J. Völk1, S. J. Wagner12, and M. Ward7

(Affiliations can be found after the references)

Received 30 June 2006 / Accepted 15 September 2006

ABSTRACT

Aims. LS 5039 is a High Mass X-ray Binary (HMXRB) comprising a compact object in an eccentric 3.9 day orbit around a massive O6.5V star.Observations at energies above 0.1 TeV (1011 eV) by the High Energy Stereoscopic System (HESS) in 2004 revealed that LS 5039 is a source ofVery High Energy (VHE) γ-rays and hence, is able to accelerate particles to multi-TeV energies. Deeper observations by HESS were carried outin 2005 in an effort to probe further the high energy astrophysics taking place. In particular, we have searched for orbital modulation of the VHEγ-ray flux, which if detected, would yield new information about the complex variation in γ-ray absorption and production within X-ray binarysystems.Methods. Observations at energies above 0.1 TeV (1011 eV), were carried out with the High Energy Stereoscopic System (HESS) of CherenkovTelescopes in 2005. A timing analysis was performed on the dataset employing the Lomb-Scargle and Normalised Rayleigh statistics, and orbitalphase-resolved energy spectra were obtained.Results. The timing analysis reveals a highly significant (post-trial chance probability <10−15) peak in the TeV emission periodogram at afrequency matching that of the 3.9 day orbital motion of the compact object around the massive stellar companion. This is the first time in γ-rayastronomy that orbital modulation has been observed, and periodicity clearly established using ground-based γ-ray detectors. The γ-ray emissionis largely confined to half of the orbit, peaking around the inferior conjunction epoch of the compact object. Around this epoch, there is also ahardening of the energy spectrum in the energy range between 0.2 TeV and a few TeV.Conclusions. The γ-ray flux vs. orbital phase profile suggests the presence of γ-ray absorption via pair production, which would imply that alarge fraction of the γ-ray production region is situated within ∼1 AU of the compact object. This source size constraint can be compared to thecollimated outflows or jets observed in LS 5039 resolved down to scales of a few AU. The spectral hardening is however not explained exclusivelyby the absorption effect, indicating that other effects are present, perhaps related to the γ-ray production mechanism(s). If the γ-ray emissionarises from accelerated electrons, the hardening may arise from variations with phase in the maximum electron energies, the dominant radiativemechanism, and/or the angular dependence in the inverse-Compton scattering cross-section. Overall, these results provide new insights into thecompeting γ-ray absorption and production processes in X-ray binaries.

Key words. gamma rays: observations – acceleration of particles – block hole physics – stars: binaries: close

1. IntroductionX-ray binaries (XRBs) comprise a compact object such as a neu-tron star or black hole in orbit with a companion star. They are

� e-mail: denauroi@in2p3.fr�� Present address: School of Chemistry and Physics, University ofAdelaide, 5005, Australia.e-mail: growell@physics.adelaide.edu.au

one of several types of astrophysical system that can providea periodic environment for the acceleration of particles and sub-sequent production of radiation. Modulation of this radiation,linked to the orbital motion of the binary system, provides keyinsights into the nature and location of particle accelerationand emission processes. While such modulation is often foundin XRBs up to hard X-ray energies (Lewin 1995; Wen et al.2006), until now it has not been established in any astrophysical

2 F. Aharonian et al.: Orbital modulation in LS 5039

source at γ-ray energies. LS 5039 (distance d ∼ 2.5 kpc) is aHMXRB comprising a compact object in a ∼3.9 day orbit arounda massive O6.5V star (Casares et al. 2005). Persistent radio out-flows (observed with extension in the range 2 to ∼1000 AU)are attributed to a mildly relativistic (v ∼ 0.2c) jet (Paredeset al. 2000, 2002), which would place LS 5039 in the micro-quasar class. Microquasars are Galactic, scaled-down versionsof Active Galactic Nuclei (AGN) (Mirabel 1994), and are asub-class of XRBs. The detection of radio (Martí et al. 1998;Ribó et al. 1999) and X-ray (Bosch-Ramon et al. 2005) emis-sion and their possible association with the MeV to GeV γ-raysources GRO J1823−12 (Collmar 2003) and 3EG 1824−1514(Paredes et al. 2000) suggests the presence of multi-GeV parti-cles. Observations in 2004 (Aharonian et al. 2005b) (∼11 h) withHESS established a new VHE γ-ray source, HESS J1826−148,within 30 arcsec of the radio position of LS 5039, revealing forthe first time that LS 5039, and hence XRBs, are capable ofmulti-TeV (1012 eV) particle acceleration. The limited statisticsdid not allow for detailed timing or variability analyses. We notethat evidence for variability at VHE γ-ray energies has recentlybeen unveiled (Albert et al. 2006) in a similar type of binary sys-tem, LS I +61◦303. Here we report on new, deeper HESS obser-vations of LS 5039 at TeV γ-ray energies, revealing that its VHEγ-ray emission is modulated by the orbital motion of the com-pact object around its massive stellar companion.

2. Observations

The observations were taken with HESS (Aharonian et al.2006b), an array of four identical Atmospheric CherenkovTelescopes (ACT) located in the Southern Hemisphere(Namibia, 1800 m a.s.l), and is sensitive to γ-rays above 0.1 TeV.The 2004 HESS observations (Aharonian et al. 2005b) havebeen followed up with a deeper campaign in 2005. After dataquality selection, the total dataset comprises 160 runs (or point-ings) representing 69.2 h observations from both 2004 and 2005.Data were analysed, employing two separate calibration pro-cedures (Aharonian et al. 2004) and several background rejec-tion and direction reconstruction methods. The results presentedhere are based on the combination of a semi-analytical showermodel and a parametrisation based on the moment method ofHillas to yield the combined likelihood of the event being ini-tiated by a γ-ray primary (de Naurois et al. 2003). As we showlater, a pure Hillas-based analysis, described in Aharonian et al.(2006b), also yielded consistent results.

3. Results

A total of 1960 γ-ray events (with an excess significance abovethe background exceeding +40σ) within 0.1◦ of the VLBA ra-dio position of LS 5039 (Ribó et al. 2002) were found. Thebest-fit position (in Galactic coordinates) is l = 16.879◦, b =−1.285◦ with statistical and systematic uncertainties of ±12 and±20 arcsec, respectively, which is consistent with the VLBA po-sition within the 1σ statistical uncertainty.

3.1. Timing analysis

A search for periodicity, by decomposing the runwiseVHE γ-ray flux at energies >1 TeV into its frequency compo-nents, was carried out using the Lomb-Scargle Test (Scargle1982), (Fig. 1) and Normalised Rayleigh Statistic (NRS)(de Jager 1994) (Fig. 2) which are appropriate for unevenly

]-1Frequency [day0 0.1 0.2 0.3 0.4 0.5 0.6

(Pro

bab

ility

)10

-lo

g

0

2

4

6

8

10

12

14

16

18

20

0

2

4

0

2

4

LS 5039, Orbital period subtracted

HESS J1825-137

LS 5039

]-1Frequency [day0.245 0.25 0.255 0.26 0.265 0.27

(Pro

bab

ility

)10

-lo

g

0

5

10

15

20

Fig. 1. Top: Lomb-Scargle (LS) periodogram of the VHE runwise fluxfor LS 5039 (chance probability to obtain the LS power vs. frequency).Inset: zoom around the highest peak (pre-trial probability ∼10−20),which corresponds to a period of 3.9078± 0.0015 days, compatiblewith the ephemeris value of 3.90603 ± 0.00017 days (vertical red lineat 0.2560 days−1 on the inset) from Casares et al. (2005). The post-trial chance probability of the orbital period peak is found to be lessthan 10−15 (see Fig. 3). Middle: LS periodogram of the same data af-ter subtraction of a pure sinusoidal component at the orbital periodof 3.90603 days. The orbital frequency peak has been removed as ex-pected, as well as significant satellite peaks (see text). Bottom: LS pe-riodogram of the HESS source HESS J1825−137 observed simultane-ously in the same field of view. The middle and bottom panel results areconsistent with that expected of white noise over the range of frequen-cies sampled. The dotted lines correspond to a 10−2 pre-trial chanceprobability.

Fig. 2. Top: normalised Rayleigh Statistic (deJager 1994) periodogramcalculated from run-wise HESS fluxes for LS 5039. The middle andbottom panels depict the NRS after subtraction of the orbital periodand for HESS J1825−137, respectively (as for the Lomb-Scargle test inFig. 1).

sampled datasets typical of those taken by HESS. The 2005 ob-servations were taken over a wide range of zenith angles yield-ing a varying energy threshold in the range 0.2 to ∼1 TeV. Toreduce adverse affects of this varying threshold in our timinganalysis, we used all events and extracted the flux normalisationabove 1 TeV assuming an average photon power-law index de-rived from all data (Γ = 2.23 for dN/dE ∼ E−Γ). Although aswe see later the photon spectral index was found to vary withinthe orbital period, the average index assumption in this methodcontributes only a small error on the derived flux above 1 TeV.

F. Aharonian et al.: Orbital modulation in LS 5039 3

Lomb-Scargle Power0 5 10 15 20 25

Lomb-Scargle Power0 5 10 15 20 25

Pro

bab

ility

-1110

-1010

-910

-810

-710

-610

-510

-410

-310

-210

-110

1

Monte-carlo normalisation

Resampling normalisation

Fig. 3. Density function of the chance probability of the Lomb-Scarglepower determined by Monte-Carlo and resampling methods after sub-traction of the orbital period sinusoid. The expected exponential densityfunctions are also indicated (solid lines). In the Monte-Carlo method,∼106 random time series were generated to produce a distribution of thehighest Lomb-Scargle power. The highest power obtained after these it-erations reaches just above 20, well below the power of 62 obtainedfrom the unshuffled lightcurve at the orbital period. An extrapolationof the curves is therefore used to estimate the chance probability forpowers above 20.

An obvious peak in the Lomb-Scargle periodogram occursat the period 3.9078± 0.0015 days (similarly observed in theNRS test), quite consistent with the orbital period determinedby Casares et al. (2005) (3.90603 ± 0.00017 days) from radialvelocity measurements of the stellar companion. The error inthis measurement was estimated from Monte-Carlo-simulatedtime series containing a sinusoid above a random background.The peak is highly significant, with a post-trial probability ofless than 10−15 that it results from a statistical fluctuation. Thischance probability was estimated via Monte-Carlo simulation ofrandom fluxes and also random re-sampling of fluxes (Fig. 3). InFig. 1 (middle panel) we also show the effect of subtracting theorbital period, which removes numerous satellite peaks that arebeat periods of the orbital period with the various gaps presentin the HESS dataset (1-day, 28-day moon cycle, 365.25-day an-nual), that is, rational fractions of beat periods added to the or-bital period. Figure 1 (bottom panel) also includes results ob-tained on the neighbouring VHE γ-ray source HESS J1825−137(Aharonian et al. 2005a), which is in the same field of view(FoV) as LS 5039 and therefore observed simultaneously. TheHESS J1825-137 periodogram does not show statistically signif-icant peaks, demonstrating that the significant peak is genuinelyassociated with LS 5039.

The ephemeris of Casares et al. (2005), determined fromDoppler-shifted optical lines (observed in 2002 and 2003),shows the binary makeup (Fig. 4) of LS 5039 as compris-ing a compact object of mass >1.38 M�1, in an eccentrice = 0.35 orbit around a stellar companion of mass ∼20 M�(with bolometric luminosity L∗ ∼ 1039 erg s−1). The separation(centre-to-centre) between these two components varies between2.2 R∗ at periastron (φ = 0.0 with reference epoch T0 (HJD-2 400 000.5)= 51 942.59) to 4.5 R∗ at apastron (φ = 0.5), for astellar radius R∗ = 7×1011 cm. A range on the system inclinationangle of 13◦ < i < 64◦ is inferred from the binary mass func-tion, the companion rotation velocity, the lack of X-ray eclipses

1 A black hole of mass 3.7+1.3−1.0 M� for the compact object was

derived by Casares et al. (2005) under the assumption of pseudo-synchronisation of the binary components.

Fig. 4. The orbital geometry (Casares et al. 2005) viewed from directlyabove LS 5039. Shown are: phases (φ) of minimum (periastron) andmaximum (apastron) separation between the two components; epochsof superior and inferior conjunctions of the compact object represent-ing phases of co-aligment along our line-of-sight of the compact ob-ject and stellar companion. The orbit is actually inclined at an angle inthe range 13◦ < i < 64◦ with respect to the view above. VHE γ-rays(straight black lines with arrows) can be absorbed by optical photons ofenergy hνε , when their scattering angle θ exceeds zero.

(which assumes that the X-ray emission occurs very close to thecompact object) and lack of Roche lobe overflow.

The phasogram (Fig. 5 top) of integral fluxes at energiesE > 1 TeV vs. orbital phase (φ) obtained on a run-by-run basis(one data run is ∼28 min) clearly indicates that the bulk of theVHE γ-ray emission is confined to roughly half of the orbital pe-riod, covering the phase interval φ ∼ 0.45 to 0.9. The VHE fluxmaximum appears to lag somewhat behind the apastron epoch,and aligns better with inferior conjunction (φ = 0.716) of thecompact object. Inferior conjunction of the compact object oc-curs when it is lined up along our line-of-sight in front of thestellar companion. The VHE flux minimum occurs at a phaseφ ∼ 0.2, slightly further along the orbit than superior conjunc-tion (φ = 0.058), which is when the compact object is linedup behind the stellar companion. Note that the inclination upperlimit i < 64◦ implies that direct views of both compact objectand stellar companion are always available. We define here twobroad phase intervals for further study: INFC (0.45 < φ ≤ 0.9)encompassing inferior conjunction, and SUPC (φ ≤ 0.45 andφ > 0.9) likewise for superior conjunction. The phase error(∆φ = (T∆P)/P2 = 0.01) due to uncertainties in the period mea-surement ∆P = 0.00017 from Casares et al. (2005), the datasetlength T ∼ 1000 days and P = 3.9 days, appears to be negli-gible. Nevertheless, further near-future optical line observationsbracketing ours at VHE γ-ray observations would be desirable tocheck for the presence of systematic drifts in the orbital period.

3.2. Phase-resolved energy spectra

The energy spectrum of the VHE γ-ray emission, and in partic-ular how it might vary with orbital phase, is an important diag-nostic tool. The differential photon energy spectrum (see Fig. 6)

4 F. Aharonian et al.: Orbital modulation in LS 5039

Orbital Phase0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

]-1

s-2

ph

cm-1

2F

> 1

TeV

[10

0

1

2

3

4

5S

uper

ior

Con

junc

tion

Infe

rior

Con

junc

tion

Infe

rior

Con

junc

tion

Sup

erio

rC

onju

nctio

n

Apa

stro

n

Per

iast

ron

Apa

stro

n

Ph

oto

n In

dex

1.5

2

2.5

3

3.5

Orbital phase0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Su

per

ior

Co

nju

nct

ion

Infe

rio

rC

on

jun

ctio

n

Infe

rio

rC

on

jun

ctio

n

Su

per

ior

Co

nju

nct

ion

Ap

astr

on

Per

iast

ron

Ap

astr

on

Orbital phase0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

)-1

s-2

cm

-1 T

eV-1

2(1

TeV

) (1

0d

EdN

0

0.5

1

1.5

2

2.5

3

3.5

4

Su

per

ior

Co

nju

nct

ion

Infe

rio

rC

on

jun

ctio

n

Infe

rio

rC

on

jun

ctio

n

Su

per

ior

Co

nju

nct

ion

Ap

astr

on

Per

iast

ron

Ap

astr

on

Fig. 5. Top: integral γ-ray flux (F > 1 TeV) lightcurve (phasogram) ofLS 5039 from HESS data (2004 to 2005) on a run-by-run basis foldedwith the orbital ephemeris of Casares et al. (2005). Each run is ∼28 min.Two full phase (φ) periods are shown for clarity. The blue solid arrowsdepict periastron and apastron. The thin red dashed lines represent thesuperior and inferior conjunctions of the compact object, and the thickred dashed line depicts the Lomb-Scargle Sine coefficients for the pe-riod giving the highest Lomb-Scargle power. This coefficient is sub-tracted from the light curve in Fig. 1 middle panel. Middle: fitted purepower-law photon index (for energies 0.2 to 5 TeV) vs. phase intervalof width ∆φ = 0.1. Because of low statistics in each bin, more compli-cated functions such as a power-law with exponential cutoff provide ano better than a pure power-law. Bottom: power-law normalisation (at1 TeV) vs. phase interval of width ∆φ = 0.1.

(0.2 to 10.0 TeV) for INFC is consistent with a hard power-lawwhere Γ = 1.85±0.06stat±0.1syst with exponential cutoff at Eo =

8.7 ± 2.0 TeV (for fitted function dN/dE ∼ E−Γ exp(−E/Eo)).In contrast, the spectrum for SUPC is consistent with a rela-tively steep (Γ = 2.53 ± 0.07stat ± 0.1syst) pure power-law (0.2 to10 TeV) (see Fig. 6). The spectra from these phase intervals aremutually incompatible, with the probability that the same spec-tral shape would fit both simultaneously being ∼2×10−6. Fittinga pure power-law (which is statistically sufficient at present) tonarrower phase intervals of width ∆φ = 0.1, and restricting thefit to energies E ≤ 5 TeV to reduce the effect of any cutoff,also demonstrates that a harder spectrum occurs when the fluxis higher (Fig. 5 middle and bottom panels). Notably, the VHEflux at E ∼ 0.2 TeV appears to be quite stable over phases andthe strongest modulation occurs at a few TeV.

E (eV)

1110 1210 1310 1410

)-1

s-2

F(E

) (e

rg c

m×2

E

-1310

-1210

-1110INFC

0.9≤ φ0.45 <

SUPC > 0.9φ 0.45 and ≤ φ

Fig. 6. Very high energy γ-ray spectra of LS 5039 for two broad orbitalphase intervals (defined in the text): INFC 0.45 < φ ≤ 0.9 (red circles);SUPC φ ≤ 0.45 and φ > 0.9 (blue triangles). The shaded regions rep-resent the 1σ confidence bands on the fitted functions (Table 1). Bothspectra are mutually incompatible with the probability that the samespectral shape would fit both simultaneously being ∼2 × 10−6. A clearspectral hardening in the region 0.3 to ∼20 TeV is noticed for the INFCphase interval.

MJD53100 53200 53300 53400 53500 53600 53700

]-1

s-2

ph

. cm

-12

F >

1 T

eV [

10

0

1

2

3

4

5

MJD53100 53200 53300 53400 53500 53600 53700

]-1

s-2

ph

. cm

-12

F >

1 T

eV [

10∆ -1

0

1

2

3

4

2004 2005

Fig. 7. Top: integral flux (E > 1 TeV) vs. time (MJD) for LS 5039 ona run-by-run basis. Bottom: after subtraction of the orbital period of3.9063 days (this is achieved by subtraction of the Lomb-Scargle co-efficients for the selected period). The average flux (dashed lines) forthe post-subtracted light-curve is consistent with a steady source with achance probability of 4×10−2. In addition, the average flux levels deter-mined exclusively for 2004 and 2005 data differ at a 2σ level (statisticalerrors) but are consistent when factoring in systematic errors of ∼10%on a run-by-run basis.

We found no evidence for long-term secular variations inVHE flux on a yearly scale independent of the orbital modu-lation (Fig. 7). The orbital modulation represents a VHE γ-rayluminosity (0.2 to 10 TeV; at 2.5 kpc) variation between 4 to10 × 1033 erg s−1. Spectral fits and other numerical results aresummarised in Table 1. Finally, in Fig. 8 we show that the spec-tral results obtained from a pure Hillas-based analysis using aseparate calibration procedure are consistent with results fromthe Semi-analytical Model+Hillas results within statistical er-rors.

4. Discussion

The basic paradigm of VHE γ-ray production requires the pres-ence of particles accelerated to multi-TeV energies and a target

F. Aharonian et al.: Orbital modulation in LS 5039 5

Table 1. Photon energy spectra and luminosity (0.2 to 10 TeV) of the VHE γ-ray emission of LS 5039 for different orbital phase intervals.The broad phase intervals INFC and SUPC encompass the inferior and superior conjunction epochs. The orbital phase φ is calculated from theephemeris of Casares et al. (2005). The best-fit function is indicated for each phase interval. The errors quoted are statistical with systematic errorsin the 10 to 15% range.

Orbital phase interval N Γ Eo Luminosity†×10−12 [ph cm−2 s−1 TeV−1] [TeV] [erg s−1]

Fit function: dN/dE = NE−Γ exp (−E/Eo)

INFC (0.45 < φ ≤ 0.9) 2.28 ± 0.10 1.85 ± 0.06 8.7 ± 2.0 1.1 × 1034

SUPC (φ ≤ 0.45) and φ > 0.9 0.91 ± 0.07 2.53 ± 0.07 – 4.2 × 1033

All Phases (time averaged) 1.85 ± 0.06 2.06 ± 0.05 13.0 ± 4.1 7.8 × 1033

Power-Law fit for energies E ≤ 5 TeV (narrow phase intervals)

0.0–0.1 0.84 ± 0.17 2.62 ± 0.230.1–0.2 0.46 ± 0.21 3.08 ± 0.470.2–0.3 0.81 ± 0.13 2.46 ± 0.200.3–0.4 0.78 ± 0.18 2.77 ± 0.600.4–0.5 1.88 ± 0.27 2.32 ± 0.170.5–0.6 2.64 ± 0.20 2.04 ± 0.100.6–0.7 2.20 ± 0.23 1.88 ± 0.130.7–0.8 2.44 ± 0.20 1.90 ± 0.100.8–0.9 2.96 ± 0.17 1.94 ± 0.080.9–1.0 1.21 ± 0.25 1.84 ± 0.25

† At 2.5 kpc distance.

E (eV)

1110 1210 1310 1410

)-1

s-2

F(E

) (e

rg c

m×2

E

-1310

-1210

-1110INFC

0.9≤ φ0.45 <

SUPC > 0.9φ 0.45 and ≤ φ

Fig. 8. Energy spectra of LS 5039 separated into the same broad phaseintervals as for Fig. 6. INFC 0.45 < φ ≤ 0.9 (red circles); SUPCφ ≤ 0.45 and φ > 0.9 (blue triangles), comparing results from the semi-analytical Model+Hillas (filled markers) and Hillas-only (open mark-ers) analyses (Aharonian et al. 2006b), respectively. The Hillas-onlyanalysis has made use of an independent calibration chain.

comprising photons and/or matter of sufficient density. In mi-croquasars, particle acceleration could take place directly insideand along the jet, out to parsec-scale distances, and also at jet ter-mination regions due to interaction with ambient matter (Heinz& Sunyaev 2002). A non-jet scenario based on acceleration inshocks created by the interaction of a pulsar wind with the windof the stellar companion has also been suggested (Maraschi &Treves 1981; Tavani et al. 1997; Dubus 2006b). The nature of theparent particles responsible for VHE γ-ray emission is under the-oretical discussion, with both accelerated electron (Aharonian &Atoyan 1998; Paredes et al. 2006) and hadron (Distefano et al.2002; Romero et al. 2003) scenarios proposed. Observationally,electrons (e.g. Corbel et al. 2002; Angelini et al. 2003) andhadrons (Margon 1984) are both known to be present inside jets.

The orbital modulation in LS 5039, with a peak fluxaround inferior conjunction, minimum flux around superior

conjunction, and hardening of energy spectra, provides new in-formation about the physical processes in microquasars. Our re-sults can be compared with those at X-ray energies (3–30 keV),where interestingly, a spectral hardening with flux is also ob-served, as well as an indication for higher fluxes at φ ∼ 0.8 in thephase-resolved light curve (Bosch-Ramon et al. 2005). The VHEγ-ray modulation is an unambiguous sign that periodic changesin the VHE γ-ray absorption and/or production processes are oc-curring, and we discuss briefly how these could arise along withissues concerning the location and size of the γ-ray productionregion.

4.1. Gamma-ray absorption

VHE γ-rays produced close enough to the stellar companionwill unavoidably suffer severe absorption via pair production(e+e−) on its intense optical photon field. The cross-sectionfor pair production is dependent on the angle θ between theVHE γ-ray and optical photons, and has an energy thresholdvarying as 1/(1 − cos θ). The level of absorption therefore de-pends on alignment between the VHE γ-ray production region,the star, and the observer, leading to orbital modulation of theVHE γ-ray flux (Protheroe & Stanev 1987; Moskalenko 1995;Böttcher & Dermer 2005; Dubus 2006a). For the orbital ge-ometry of LS 5039, in which apastron and periastron are nearthe inferior and superior conjunction epochs, the absorption ef-fect provides flux maxima at the phase of inferior conjunction,when γ photons are emitted towards the observer parallel to stel-lar photons (cos θ = 1), and flux minima at superior conjunction(Dubus 2006a). This picture agrees well with the observed pha-sogram (Fig. 5), suggesting that absorption plays an importantrole. An additional key expectation is that the strongest absorp-tion, and hence modulation, should occur in the energy rangeE ∼ 0.2 to 2 TeV (Dubus 2006a). However, the observationsshow that the flux at ∼0.2 TeV appears quite stable, suggestingthat additional processes must be considered to explain the spec-tral modulation.

6 F. Aharonian et al.: Orbital modulation in LS 5039

4.2. Gamma-ray production

VHE γ-ray emission can be produced by accelerated electronsthrough the inverse-Compton (IC) scattering of stellar photonsof the companion star, and/or from accelerated hadrons throughtheir interaction with surrounding photons and particles. The ef-ficiency of VHE γ-ray production under these scenarios willpeak around periastron (φ = 0.0), reflecting the minimal sepa-ration between particle acceleration sites and targets, and highertarget photon densities. The observed phasogram is in contrastwith this behaviour. However, important influences on the en-ergy spectrum can arise from variations of the maximum en-ergy of accelerated electrons and dominance in the radiativeprocesses (IC and/or synchrotron emission) by which they loseenergy (known as cooling). The high temperature of the com-panion star means that IC γ-ray production proceeds primarilyin the deep Klein-Nishina regime (where the IC cross-section issharply reduced compared to the Thompson regime). The max-imum electron energy Emax is derived by equating the com-peting acceleration and dominant radiative loss timescales. Thehard VHE γ-ray spectrum (Γ ≤ 2.5) implies IC is the domi-nant cooling mechanism. Above some high energy ε (discussedshortly below), synchrotron losses however are expected to takeover. For dominant IC cooling we have (Aharonian et al. 2006a)Emax ∝ (B/w)3.3, where B is the magnetic field and w is the targetphoton energy density. Since w ∝ 1/d2 (for binary separation d),and assuming B ∝ 1/d, Emax will increase by a factor ∼10 fromperiastron to apastron, leading to a spectral hardening around theapastron phase (for LS 5039, also near inferior conjunction). Formagnetic fields expected within the binary orbit, B ∼ 1 G, elec-trons with energy above ε ≈ 6[(B/G)(d/R∗)]−1 TeV2 will coolvia synchrotron (X-ray) radiation in preference to IC cooling.Much stronger synchrotron losses will then produce a steepen-ing of the VHE γ-ray spectrum (Moderski et al. 2005) and thusthe d and B dependence of this changeover energy ε could alsointroduce a spectral hardening at apastron. Spectral hardeningcould also come from the angular dependence of the IC scatter-ing cross-section (Khangulyan & Aharonian 2005), which peaksat small scattering angles (i.e. around inferior conjunction). It isalso feasible to consider production via accelerated hadrons. Forconsiderably larger magnetic fields, for example at the base ofthe jet where B ∼ 105 G could be expected, protons can be ac-celerated to multi-TeV energies. In such high magnetic fields, thestrong cooling of electrons will also render any IC componentsnegligible (see for example Aharonian et al. 2006a). Acceleratedprotons interacting with stellar wind particles and possibly X-rayphotons3 associated with an accretion disk, provide a plausiblehadronic origin via the decay of neutral pions.

4.3. Pair cascades

For magnetic fields B < 10 G (Aharonian et al. 2006a),an extra complication arises from pair cascades initiated byabsorption of first generation VHE γ-rays (Aharonian et al.2006a; Bednarek 2006). Cascades significantly increase thetransparency of the source at ∼TeV energies, reducing absorp-tion effects. Interestingly we do observe a VHE signal (excesssignificance 6.1σ; 79 γ-rays) in a phase interval φ = 0.0 ± 0.05where VHE γ-rays are unlikely to escape due to absorption

2 The electron energy at which synchrotron and IC cooling times areequal.

3 The X-ray disk luminosity, ∼1034 erg s−1, is marginally sufficientunder this process to meet by itself that of the VHE γ-ray emission.

(Dubus 2006a). A cascade, or at least an additional, possibly un-modulated component from another process may be present.

4.4. γ-ray production region: location and size

If the phasogram profile is the result of absorption, the γ-rayproduction region, or at least a large fraction of it, should be em-bedded within the stellar photosphere defined where the absorp-tion optical depth τγγ is ≥1. Such an optical depth occurs within∼1 AU of the stellar companion. However this size constraintcomes with a caveat since absorption is unlikely to be the soleprocess present as argued earlier. An unmodulated component,possibly explaining the φ = 0.0 ± 0.05 signal could for examplearise if parts of the source lay outside the stellar photosphere,or is always situated in front of the stellar companion along theline of sight. The size constraint is considerably smaller thanthe jets/outflows observed out to a distance of ∼1000 AU fromthe binary system (Paredes et al. 2002), and similar in size to thesmaller-scale jets/outflows observed out to ∼2 AU (Paredes et al.2000). The size constraint is also significantly smaller than the∼0.3 pc (28 arcsec at 2.5 kpc) upper limit (1σ) on the sourceradius based on the HESS angular resolution.

5. ConclusionsIn conclusion, new observations by HESS have established or-bital modulation of the VHE γ-ray flux and energy spectrumfrom the XRB LS 5039. The flux vs. orbital phase profile pro-vides the first indication for γ-ray absorption within an astro-physical source, suggesting that a large part of the VHE γ-rayproduction region lies inside the photosphere (within ∼1 AU)of the massive stellar companion. However, not all of the ob-served effects can be explained by absorption alone. Modulationof the energy spectrum could arise from changes in the maxi-mum energies of electrons responsible for the radiation, changesin the dominant radiative mechanism, and/or scattering angledependence of the inverse-Compton scattering effect. A VHEγ-ray signal near φ = 0.0 may arise from pair-cascades, oran unmodulated component produced outside the photosphere.These observations provide key information about the astro-physics associated with particle acceleration processes and sub-sequent VHE γ-ray production in XRBs. In particular, we arenow able to begin to explore in detail the complex relationshipbetween γ-ray absorption and production processes within thesebinary systems.

Acknowledgements. The support of the Namibian authorities and of theUniversity of Namibia in facilitating the construction and operation of HESS isgratefully acknowledged, as is the support of the German Ministry for Educationand Research (BMBF), the Max Planck Society, the French Ministry forResearch, the CNRS-IN2P3 and the Astroparticle Interdisciplinary Programmeof the CNRS, the UK Particle Physics and Astronomy Research Council(PPARC), the IPNP of Charles University, the South African Department ofScience and Technology and National Research Foundation, and the Universityof Namibia. We appreciate the excellent work of the technical support staff inBerlin, Durham, Hamburg, Heidelberg, Palaiseau, Paris, Saclay, and Namibia inthe construction and operation of the equipment. L.C., C.M., M.O., and M.T.are also affiliated with the European Associated Laboratory for Gamma-RayAstronomy, jointly supported by CNRS and the Max Planck Society.

References

Aharonian, F., & Ayotan, A. 1998, New Astron. Rev., 42, 579Aharonian, F., Akhperjanian, A. G., Aye, K.-M., et al. (HESS Collaboration)

2004, Astropart. Phys., 22, 109Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R., et al. (HESS

Collaboration) 2005a, A&A, 442, L25Aharonian, F., Akhperjanian, A. G., Aye, K.-M., et al. (HESS Collaboration)

2005b, Science, 309, 746

F. Aharonian et al.: Orbital modulation in LS 5039 7

Aharonian, F., Anchordogui, L., Khangulyan, D., & Montaruli, T. 2006a, J. Phys.Conf. Ser., 39, 408

Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R., et al. (HESSCollaboration) 2006b, A&A, 457, 899

Albert, J., Aliu, E., Anderhub, H., et al. (MAGIC Collaboration) 2006, Science,312, 1771

Angelini, L., White, N., & Ball, L. 2003, ApJ, 586, L71Bednarek, W. 2006, MNRAS, 368, 579Bosch-Ramon, V., Paredes, J. M., Ribó, M., et al. 2005, ApJ, 628, 388Böttcher, M., & Dermer, C. D. 2005, ApJ, 634, L81Casares, J., Ribó, M., Ribas, I., et al. 2005, MNRAS, 364, 899Collmar, W. 2003, Proc. 4th Agile Science Workshop, Frascati (Rome), 177Corbel, S., Fender, R. P., Tzioumis, A. K., et al. 2002, Science, 298, 196de Jager, O. C. 1994, ApJ, 436, 239de Naurois, M. 2005, Proc. Towards a Network of Atmospheric Cherenkov

Detectors VII (Palaiseau) 149 [arXiv:astro-ph/0607247]Distefano, C., Guetta, D., Waxman, E., & Levinson, A. 2002, ApJ, 575, 378Dubus, G. 2006a, A&A, 451, 9Dubus, G. 2006b, A&A, 456, 801Harmon, B. A., Wilson, C. A., Fishman, G. J., et al. 2004, ApJS, 154, 585Hartman, R. C., Bertch, D. L., Bloom, S. D., et al. 1999, ApJS, 123, 79Heinz, S., & Sunyaev, R. 2002, A&A, 390, 751Khangulyan, D., & Aharonian, F. 2005, AIP Conf. Proc., 745, 359Lewin, W. H. G., et al., X-Ray Binaries, Cambridge Astrophys. Ser. (Cambridge

University Press 1995)Maraschi, L., & Treves, A. 1981, MNRAS, 194, 1Margon, B. 1984, ARA&A, 22, 507Martí, J., Paredes, J. M., & Ribó, M. 1998, A&A, 338, L71Mirabel, I. F., & Rodriguez, L. F. 1994, Nature, 371, 46Moderski, R., Sikora, M., Coppi, P. S., & Aharonian, F. 2005, MNRAS, 363, 954Moskalenko, I. V. 1995, Space Sci. Rev., 72, 593Paredes, J. M., Martí, J., Ribó, M., & Massi, M. 2000, Science, 288, 2340Paredes, J. M., Ribó, M., Ros, E., & Martí, J. 2002, A&A, 393, L99Paredes, J. M., Bosch-Ramon, V., & Romero, G. E. 2006, A&A, 451, 259Protheroe, R. J., & Stanev, T. 1987, ApJ, 322, 838Ribó, M., Paredes, J. M., & Martí, M. 1999, New Astron. Rev., 43, 545Ribó, M., Paredes, J. M., Romero, G. E., et al. 2002, A&A, 384, 954Romero, G. E., Torres, D. F., Kaufman Bernadó, M. M., & Mirabel, I. F. 2003,

A&A, 410, L1Scargle, J. D. 1982, ApJ, 263, 835Tavani, M., & Arons, J. 1997, ApJ, 477, 439Wen, L., Levine, A. M., Corbet, R. H. D., & Bradt, H. V. 2006, ApJS, 163, 372

1 Max-Planck-Institut für Kernphysik, PO Box 103980, 69029Heidelberg, Germany

2 Yerevan Physics Institute, 2 Alikhanian Brothers St., 375036Yerevan, Armenia

3 Centre d’Étude Spatiale des Rayonnements, CNRS/UPS, 9 Av. duColonel Roche, BP 4346, 31029 Toulouse Cedex 4, France

4 Universität Hamburg, Institut für Experimentalphysik, LuruperChaussee 149, 22761 Hamburg, Germany

5 Institut für Physik, Humboldt-Universität zu Berlin, Newtonstr. 15,12489 Berlin, Germany

6 LUTH, UMR 8102 du CNRS, Observatoire de Paris, Section deMeudon, 92195 Meudon Cedex, France

7 University of Durham, Department of Physics, South Road, DurhamDH1 3LE, UK

8 Unit for Space Physics, North-West University, Potchefstroom2520, South Africa

9 Laboratoire Leprince-Ringuet, IN2P3/CNRS, École Polytechnique,91128 Palaiseau, France

10 APC, 11 place Marcelin Berthelot, 75231 Paris Cedex 05, FranceUMR 7164 (CNRS, Université Paris VII, CEA, Observatoire deParis).

11 Dublin Institute for Advanced Studies, 5 Merrion Square, Dublin 2,Ireland

12 Landessternwarte, Universität Heidelberg, Königstuhl, 69117Heidelberg, Germany

13 Laboratoire de Physique Théorique et Astroparticules,IN2P3/CNRS, Université Montpellier II, CC 70, Place EugèneBataillon, 34095 Montpellier Cedex 5, France

14 DAPNIA/DSM/CEA, CE Saclay, 91191 Gif-sur-Yvette Cedex,France

15 Laboratoire d’Astrophysique de Grenoble, INSU/CNRS, UniversitéJoseph Fourier, BP 53, 38041 Grenoble Cedex 9, France

16 Institut für Astronomie und Astrophysik, Universität Tübingen,Sand 1, 72076 Tübingen, Germany

17 Laboratoire de Physique Nucléaire et de Hautes Énergies,IN2P3/CNRS, Universités Paris VI & VII, 4 place Jussieu, 75252Paris Cedex 5, France

18 Institute of Particle and Nuclear Physics, Charles University,V Holesovickach 2, 180 00 Prague 8, Czech Republic

19 Institut für Theoretische Physik, Lehrstuhl IV: Weltraum undAstrophysik, Ruhr-Universität Bochum, 44780 Bochum, Germany

20 University of Namibia, Private Bag 13301, Windhoek, Namibia21 Universität Erlangen-Nürnberg, Physikalisches Institut, Erwin-

Rommel-Str. 1, 91058 Erlangen, Germany22 European Associated Laboratory for Gamma-Ray Astronomy,

jointly supported by CNRS and MPG